Discovery of Hypoxanthine and Inosine as Robust Biomarkers for Predicting the Preanalytical Quality of Human Plasma and Serum for Metabolomics

Abstract

In cold human blood, the anomalous dynamics of adenosine triphosphate (ATP) result in the progressive accumulation of adenosine diphosphate (ADP), adenosine monophosphate (AMP), inosine monophosphate (IMP), inosine, and hypoxanthine. While the ATP, ADP, AMP, and IMP are confined to red blood cells (RBCs), inosine and hypoxanthine are excreted into plasma/serum. The plasma/serum levels of inosine and hypoxanthine depend on the temperature of blood and the plasma/serum contact time with the RBCs and hence they represent robust biomarkers for evaluating the preanalytical quality of plasma/serum. These biomarkers are highly specific since they are generally not present or at very low levels in fresh plasma/serum and are highly sensitive since they are derived from ATP, one of the most abundant metabolites in blood. Further, whether or not blood was kept at room temperature or on ice could be predicted based on inosine levels. An analysis of >2000 plasma/serum samples processed for metabolomics-centric analyses showed alarmingly high levels of inosine and hypoxanthine. The results highlight the gravity of sample quality challenges with high risk of grossly inaccurate measurements and incorrect study outcomes. The discovery of these robust biomarkers provides new ways to address the longstanding and underappreciated preanalytical sample quality challenges in the blood metabolomics field.

Graphical Abstract

INTRODUCTION

The field of metabolomics deals with the analysis of small molecule metabolites in biological specimens and provides information on functions of genes, enzymes, and transcripts in human health and diseases as well as effects of drugs, environmental toxicants, food, and nutrition. Blood is the most widely used biospecimen in the metabolomics field due to its clinical relevance that arises from its association with every living cell in the human body and its relatively easy access for routine investigations. Human blood consists of nearly a 50:50 (v/v) mixture of plasma and blood cells. Blood cells include red blood cells (RBCs), white blood cells, and platelets, of which RBCs alone constitute >99%. Blood metabolomics typically refers to the study of metabolites of plasma or serum, which is obtained by separating cells from whole blood. To ensure plasma metabolite profiles are not contaminated or confounded by the undesired effect of metabolism in blood cells, it is critical that the blood is processed as soon as it is drawn.¹ However, this requirement is stringent; it is generally challenging to achieve in routine applications owing to logistical considerations.

For several decades, numerous investigations have focused on circumventing this challenge and establishing optimized preanalytical conditions to harvest plasma/serum from whole blood.^2–8 Each study has utilized different experimental strategies and analytical platforms and focused on different numbers and classes of metabolites. All reported altered plasma metabolite levels as a function of blood storage time and temperature. Specific inferences of these studies, however, were varied and depended on the metabolites evaluated. A consistent finding was that metabolite levels change more rapidly when blood is stored on the bench at room temperature compared to when it is stored on ice.^5,6,9–15 Some studies, however, reported contradictory^16–20 or even misleading results that blood metabolites are stable for up to 24 hrs.¹⁶ The current consensus is that the integrity of the plasma metabolite profile does not alter appreciably when blood is stored on ice and processed within a few hours (typically < 4 hrs.).^6,7 However, such a norm is vague and lacks the ability to objectively validate or predict the preanalytical quality of plasma in routine studies. This is a greatly underappreciated challenge, which carries the risk of grossly inaccurate results and incorrect study outcomes. Robust biomarkers are therefore critically needed to assess the preanalytical quality of plasma/serum samples for metabolomics studies. To this end, numerous studies have identified some potential metabolite biomarkers. However, no biomarkers have gained widespread use to date. For example, a study using liquid chromatography mass spectrometry (LC-MS) detected increases in hypoxanthine and sphingosine phosphate levels as a function of time the blood was exposed to room temperature. However, it suggested that plasma metabolites are stable for up to 4 hrs. when blood is immediately placed on ice.^6,7 Another study using gas chromatography (GC-MS) and LC-MS detected 45 plasma metabolites including hypoxanthine, betaine, kynurenine, pyruvic acid, and different lipid types that distinguish preanalytical conditions.¹¹ A few recent studies have incorporated modeling of the data acquired by NMR spectroscopy¹³ or LC-MS²¹ to predict pre-centrifugation temperature and times. These studies detected many identified and unidentified metabolites including hypoxanthine, pyruvate, lactate and lipids. However, the lack of robust biomarkers that can be consistently detected by both NMR and MS with high sensitivity and specificity continues to deleteriously affect the critical evaluation of the preanalytical sample quality in blood metabolomics.

NMR spectroscopy exhibits numerous unparalleled characteristics including the high reproducibility, quantitative ability, and the capability to develop methods for analysis of unstable metabolites such as redox coenzymes, energy coenzymes, and antioxidants that are fundamental to the cellular functions.^22–25 Using NMR, we recently discovered anomalous dynamics of the redox coenzymes and energy coenzymes, whose levels are altered massively even in cold blood stored for a short time (< 30 min).²⁶ These results are surprising and contrary to the current knowledge about metabolite stability and provide compelling evidence that warrants reconsidering blood collection protocols for reliable analysis of blood metabolites. In the current study, investigation on the effect of the anomalous dynamics of metabolites in blood showed that, in cold blood, the major energy coenzyme, adenosine triphosphate (ATP) causes the accumulation of adenosine diphosphate (ADP), adenosine monophosphate (AMP), inosine monophosphate (IMP), inosine, and hypoxanthine. While the ATP, ADP, AMP, and IMP are confined to the RBC, inosine and hypoxanthine are excreted into plasma. Plasma hypoxanthine and inosine levels depend on the blood storage temperature and the plasma contact time with the RBC. Based on these results, we demonstrate that hypoxanthine and inosine represent robust biomarkers for validating as well as predicting the preanalytical quality of plasma. This discovery is significant considering the lack of objective biomarkers to date has deleteriously affected metabolomics applications of plasma/plasma, a widely used biospecimen in the field.

MATERIALS AND METHODS

Chemicals and solvents:

Methanol, chloroform, sodium phosphate (monobasic; NaH₂PO₄), sodium phosphate (dibasic; Na₂HPO₄), fumaric acid, maleic acid, and 3-(trimethylsilyl)propionic acid-2,2,3,3-d₄ sodium salt (TSP) were obtained from Sigma-Aldrich (St. Louis, MO). Deuterium oxide (D₂O) was obtained from Cambridge Isotope laboratories, Inc. (Andover, MA). Deionized (DI) water was purified using an in-house Synergy Ultrapure Water System from Millipore (Billerica, MA). All chemicals were used with no further purification.

Blood collection and pretreatment:

Human blood from healthy individuals was collected in heparinized BD Vacutainer tubes (BioVision, CA). The biospecimen collection protocol was approved by the IRB from the University of Washington. Figure 1 describes the preanalytical treatments of whole blood and plasma specimens. Briefly, each freshly drawn blood (35 – 40 mL; n=10 individuals) was aliquoted (200 μL or 1200 μL) into Eppendorf tubes (2 mL).

Figure 1.

Each freshly drawn blood was aliquoted (200 μL or 1200 μL) into Eppendorf tubes (2 mL). The 200 μL aliquots were split equally into four groups, each group containing 7 samples. Two groups were placed on ice and the remaining two groups were placed on the bench at room temperature (RT). Blood specimens from one group that was placed on ice and another group that was placed on the bench were mixed with a mixture of methanol and chloroform (1:2:2 v/v/v). Subsequently, one solvent treated sample and one untreated sample from each group were stored at −80 °C at different time points varying from 0 to 24 hrs. until used for further analysis. Separately, the 1200 μL aliquots were split into two groups. One group was placed on ice and the other at RT. One sample from each group was then centrifuged at 2000 × g at 4 °C for 15 min at different times varying from 0 hrs. to 144 hrs. The supernatant plasma (200 μL aliquots) was transferred to fresh vials and stored at −80 °C until used for further analysis.

Whole Blood:

the 200 μL aliquots were split equally into four groups, each group containing 7 samples. Two groups were placed on ice and the remaining two groups were placed on the bench at room temperature (RT). Blood specimens from one group that was placed on ice and another group that was placed on the bench were added with a mixture of methanol and chloroform (1:2:2 v/v/v). Subsequently, one solvent treated sample and one untreated sample from each group were stored at −80 °C at different time points varying from 0 to 24 hrs. until used for further analysis.

Plasma Samples:

The 1200 μL aliquots were split into two groups. One group was placed on ice and the other was placed on the bench at room temperature. One sample from each group was then centrifuged at 2000×g and 4 °C for 15 min after different delay times varying from 0 hrs. to 144 hrs., the supernatant plasma (200 μL aliquots) was transferred to fresh vials and stored at −80 °C until used for further analysis.

Serum Samples:

Separately, blood was collected in BD Vacutainer tubes (BioVision, CA) from six healthy individuals. For each individual, blood (~1.5 mL) was collected in six different tubes and kept on the bench at room temperature for blood clotting. One blood sample from each individual was then centrifuged at 2000×g and 4 °C for 15 min after different delay times varying from 0.5 hrs. to 144 hrs. The supernatant serum was transferred to fresh vials and stored at −80 °C until used for further analysis.

Commercial plasma and serum samples:

Plasma (n=1817) and serum (n=230) samples that represent eight different cohorts (cohort 1 to cohort 8) obtained from geographically distinct sites in the United States were used to evaluate the preanalytical sample quality based on the biomarkers discovered in this study (see Table S1). These samples were procured previously as part of other studies and NMR spectra had been obtained in our laboratory as described previously.²⁷ Briefly, metabolites from 200 μL plasma or serum were extracted using methanol in the ratio 1:2 (v/v), supernatants were dried, dried residues were dissolved in 200 μL phosphate buffer in D₂O containing TSP, the solutions were transferred to 3 mm NMR tubes and used for NMR analyses as described below.

Preparation of phosphate buffer:

Buffer solution (100 mM) was prepared by dissolving 1124 mg anhydrous Na₂HPO₄ and 250 mg anhydrous NaH₂PO₄ in 100 g D₂O. TSP (124 μM) and fumaric acid (98.6 μM) or maleic acid (255.5 μM) were added as internal references for the chemical shift scaling and quantitation, respectively.²⁸ The calculated pH of the buffer solution was 7.4 and the measured pH was 7.33. This buffer was used without further pH correction.

Whole blood metabolite extraction:

Frozen blood samples that had not been treated with organic solvent prior to storing at −80 °C were added with a mixture of methanol and chloroform (1:2:2 v/v/v/). No further solvent was added to the remaining samples since they were already treated with solvents prior to storing at −80 °C. Next, all samples were vortexed for 2 min or until a homogenous mixture was formed, sonicated for 20 min at 4 °C, and vortexed again for 30 s. The mixtures were then centrifuged at 13,400×g for 30 min to separate proteins/cell debris. Clear solutions were transferred to fresh vials and dried using nitrogen gas. The dried samples were mixed with 200 μL phosphate buffer in D₂O containing fumaric acid and TSP, and then transferred to 3 mm NMR tubes. The D₂O buffer was degassed using helium gas prior to mixing with samples and the NMR tubes were flushed with the same gas before and after transferring solutions and sealed with parafilm to prevent air from entering the tube.^22–24

Plasma/serum metabolite extraction:

Frozen plasma/serum (200 μL) samples were mixed with methanol (1:2 v/v/), vortexed for 30 s and stored at 20 °C for 20 min. The mixtures were then centrifuged at 13,400×g for 40 min to separate proteins. Clear solutions were transferred to fresh vials and dried using nitrogen gas. The dried samples were mixed with 200 μL phosphate buffer in D₂O containing maleic acid and/or TSP, and then transferred to 3 mm NMR tubes. Separately, for mass spectrometry analysis, the plasma samples (200 μL) were protein precipitated using the same protocol as described above and dried. The dried samples were dissolved in DI water, 10 μL of which was mixed with a 240 μL mixture of MS analysis solvents (water:acetonitrile:methanol in 4:5.7:0.25 (v/v/v) ratio) containing ammonium acetate (11.7 mM), acetic acid (0.24 %) and isotope labeled internal standards and used for targeted LC–MS/MS analysis.

Intact plasma/serum:

Frozen plasma/serum samples (100 μL) were mixed with 100 μL phosphate buffer in D₂O containing maleic acid and/or TSP and transferred to 3 mm NMR tubes.

NMR Spectroscopy:

NMR experiments were performed at 298 K on a Bruker Avance III 800 MHz spectrometer equipped with a cryogenically cooled probe and Z-gradients suitable for inverse detection. The NOESY pulse sequence with water suppression and the CPMG (Carr-Purcell-Meiboom-Gill) pulse sequence with residual water suppression using presaturation were used for ¹H 1D NMR experiments. Spectra were obtained using a 9615 Hz spectral width, 32,768 time-domain points, and 5 s recycle delay. For the CPMG experiment, a pulse train length of 128 or 256 ms was used. Separately, to confirm newly identified metabolite peaks, ¹H spectra were obtained after the addition of the stock solutions (1 to 10 μL; 1.0 or 50 mM) containing the authentic compounds, individually. For solutions of inosine and hypoxanthine standards as well as a typical plasma sample, spectra were also obtained at different recycle delay varying from 5 to 120 s to determine the delay for the maximum recovery of NMR peaks. The raw data were Fourier transformed using a spectral size of 32,768 points after multiplying by an exponential window function with a line broadening of 0.5 Hz.

To identify unknown metabolites, homonuclear two-dimensional (2D) experiments such as ¹H-¹H correlation spectroscopy (COSY) using the pulse sequence ‘cosygpqf’ and ¹H-¹H total correlation spectroscopy (TOCSY) using the pulse sequence ‘mlevphpr’ were performed for a typical plasma sample. The 2D experiments were performed with or without suppression of the residual water signal by presaturation during the relaxation delay. A sweep width of 9615 Hz was used in both dimensions; 512 FIDs were obtained with t₁ increments, each with 2048 complex data points. The number of transients used was 16 for COSY and 40 for TOCSY and the relaxation delay was 1.0 s. The resulting 2D data were zero-filled to 4096 (for COSY) or 2048 (for TOCSY) points in the t₂ dimension and 1024 points in the t₁ dimension. A 45° (for COSY) or 90° (for TOCSY) shifted squared sine-bell window function was applied to both dimensions before Fourier transformation. The spectra were phase and baseline corrected and the chemical shifts were referenced to the internal standard, TSP, signal for both 1D and 2D spectra.

Relaxation time (T₁) measurement.

NMR experiments were performed using the inversion recovery pulse sequence for a mixture of inosine (300 μM) and hypoxanthine (300 μM) separately in D₂O buffer and H₂O:D₂O (90:10 v/v) buffer. NMR spectra were obtained using 90° and 180° RF pulses, 60 s recycle delay, 10204 Hz spectral width, 65,536 time-domain points, 4 dummy scans, and 32 transients. A total of 25 spectra were obtained with the inversion recovery delay varying between 0.1 and 60 s. Based on the spectra, the inversion recovery delay for null signal (τ_null) was identified and T₁ values for inosine and hypoxanthine were calculated using the equation $T_1=\frac{{\tau }_{\mathit{\text{null}}}}{\text{ln}2}$. Bruker Topspin versions 4.1.4 and 3.6.5 software packages were used for NMR data acquisition, processing and analyses.

Spectral analysis, metabolite identification, quantitation and data analysis.

Assignment of metabolite peaks in the ¹H NMR spectra was based on the templates we established previously for blood and plasma/serum.^24,29,30 Unknown metabolites were identified based on the comprehensive analysis of 1D and 2D NMR spectra. Metabolite concentrations in time series spectra were obtained based on the integration of the characteristic peaks using the Bruker AMIX software after applying corrections for T₁ relaxation. Metabolite concentrations among different preanalytical treatments were compared. Separately, the concentrations of inosine and hypoxanthine in eight cohorts of plasma/serum samples (see Table S1) were obtained similarly using their ¹H 1D NMR spectra. The data were evaluated to identify potential biomarkers of sample quality.

Mass Spectrometry:

Plasma metabolites were analyzed by targeted liquid chromatography-mass spectrometry (LC-MS/MS) using an AB Sciex 6500+ Triple Quadrupole MS equipped with ESI ionization source as described previously.³¹ The instrument was attached to two Shimadzu UPLC pumps and the pumps were connected to an auto-sampler in parallel so that chromatography separation can be performed using two analytical hydrophilic interaction liquid chromatography (HILIC) columns, independently, one for positive ionization mode and the other for negative ionization mode. Identical columns (Waters XBridge BEH Amide XP) were used for both the LC systems and sample was injected for each column separately. While one column was performing separation and MS data acquisition in ESI+ ionization mode, the other column was equilibrated and readied for analysis in ESI- mode. Each chromatography separation and data acquisition took 18 min (36 min per sample). To assess the carryover from the columns, system performance, and data reproducibility, a blank sample, a pooled human serum sample, and a pooled study sample were run once for every 10 study samples. The LC-MS system was controlled using AB Sciex Analyst 1.6.3 software. MS data acquisition was performed in multiple-reaction-monitoring (MRM) mode. Measured MS peaks were integrated using AB Sciex MultiQuant 3.0.3 software. In total, 168 time series samples from 9 subjects were analyzed. A total of 361 aqueous metabolites were targeted and 244 were detected in each sample. One sample and three metabolites NADH, fructose, and palmitic acid were removed from analysis due to poor peak shape or interference from other metabolites such as glucose. Additionally, allopurinol was removed as it is an exogenous compound. The remaining 240 metabolites (Table S2) were used to compare changes in their levels as a function of the time and temperature prior to sample processing.

RESULTS AND DISCUSSION

¹H NMR spectra of whole blood showed peaks for stable as well as labile metabolites including the major redox coenzymes (NAD⁺, NADH, NADP⁺, NADPH), energy coenzymes (ATP, ADP, AMP), and antioxidants (GSH, GSSG) (Figure S1). Blood treated with an organic solvent showed stable levels for the redox coenzymes and energy coenzymes. However, blood not treated with an organic solvent exhibited anomalous dynamics for both redox coenzymes and energy coenzymes and the results were in accordance with those reported in our recent study.²⁶ Importantly, the metabolism of ATP leads to its hydrolysis to form ADP, which further hydrolyzes to form AMP. AMP undergoes deamination to form IMP (inosine monophosphate). IMP is dephosphorylated to form inosine, which is then converted to hypoxanthine (Scheme 1). ATP, ADP, AMP, and IMP were identified in the NMR spectra of blood previously.^24,26 However, inosine and hypoxanthine were not identified in whole blood by NMR to the best of our knowledge. We established their identity based on a comprehensive analysis of 1D and 2D NMR spectra of blood and standard compounds and by performing spiking experiments using authentic standards. Concentrations of ATP, ADP, AMP, IMP, inosine, and hypoxanthine depended on the blood storage time and temperature. Figure 2 (a–c) shows typical spectra of a blood processed immediately or after storing on ice or the bench for 24 hrs. ATP levels are stable in blood immediately treated with an organic solvent after the draw (Fig. 2a), whereas in untreated blood, ATP is converted to ADP, AMP, IMP, inosine, and hypoxanthine (Fig. 2b). Inosine levels were generally low or undetectable in blood that stayed at room temperature (Fig. 2c).

Scheme 1:

In cold human blood, the anomlous dynamics of ATP in red blood cell (RBC) leads to the progressive accumulation of adenosine diphsophate (ADP), adenosine monophosphate (AMP, inosine monophosphate (IMP), inosine, and hypoxanthine. The phosphorous containing metabolites, ATP, ADP, AMP, and IMP, are confined to the RBC due to their negative charge and high polarity, whereas the non-phosphorus metabolites, inosine and hypoxanthine, are excreted into the plasma/serum.

Figure 2:

Portions of 800 MHz ¹H NMR spectra of whole human blood and plasma. (a) whole blood treated with an organic solvent immediately after the draw; (b) untreated whole blood placed on ice for 24 hrs.; (c) untreated whole blood placed on the bench at room temperature (RT) for 24 hrs.; (d) plasma separated from blood immediately after the draw; (e) plasma separated from blood placed on ice for 24 hrs.; (f) plasma separated from blood placed on the bench at RT for 24 hrs. In (b and c), a major portion of ATP is converted to ADP, AMP, IMP, inosine and/or hypoxanthine; inosine and/or hypoxanthine are excreted into plasma (e &f); inosine levels are generally low or undetectable in blood placed on the bench at RT even after 24 hrs. (c and f).

¹H NMR spectra of plasma were devoid of peaks that arise from the cellular metabolism of RBC when blood was processed as soon as it was drawn (Fig. S2a). However, when blood stayed on ice or the bench prior to separating plasma, the plasma NMR spectra exhibited prominent additional peaks from new metabolites (Fig. S2b). Comprehensive investigation of the spectra enabled the establishment of the identity of the new peaks as hypoxanthine and inosine. For this identification effort, we used a combination of 1D and 2D NMR experiments (Figs. S3 and S4), results of our previous studies,²⁹ results of the analysis of whole blood spectra (Fig. 2b–c), and spiking experiments using authentic compounds (Fig. S5).

The results of the blood and plasma NMR analysis (Fig. 2) indicate that while phosphorus containing metabolites, ATP, ADP, AMP, and IMP, are confined to RBC owing to their negative charge and high polarity, the non-phosphorus metabolites, inosine and hypoxanthine, escape or are excreted into plasma/serum (Scheme 1) (Fig. 2d–f). Plasma/serum levels of hypoxanthine and inosine depend on the temperature of blood as well as the plasma contact time with blood cells (Fig. 3; Figs. S6–S15). The levels of hypoxanthine increased somewhat exponentially with time when blood stayed on ice and it increased linearly when blood stayed on bench (Fig. 4). Up to ~1 hr., its level in plasma was <<1 μM when the blood was on ice, whereas during the same period it was >1 μM when the blood stayed at room temperature; beyond ~1 hr., its level increased progressively (Fig. 4a,c). The increase in the levels for inosine, however, was less dramatic when blood stayed on ice and it was generally undetectable or low when blood stayed on the bench (Fig. 4b,d). These results demonstrate that hypoxanthine and inosine levels represent potential biomarkers for predicting and validating the preanalytical quality of plasma. Importantly, they are highly specific since they are generally very low in plasma (Figs. 2–4) and they are highly sensitive since they are derived from ATP, one of the five most abundant metabolites of human blood (≥ 1000 μM), the other four being glucose, 2,3-bisphosphoglycerate (2,3-BPG), glutathione, and lactate.^24,25,32 The fact that inosine levels are generally undetectable by NMR when blood stayed on the bench indicates that whether or not blood was kept at room temperature or on ice could also be predicted based on plasma inosine levels. Given that the biomarkers are derived from a highly unstable and concentrated metabolite, ATP, it is comprehensible that hypoxanthine and inosine provide sensitive and specific biomarkers of blood storage conditions prior to processing.

Figure 3:

Portions of 800 MHz ¹H NMR spectra of plasma from the same subject. Plasma was separated from blood placed on ice (a-e) or the bench at room temperature (f-j) for 0, 1.5, 3.5, 5.5, or 7.5 hrs.

Figure 4:

Plasma hypoxanthine and inosine levels when blood is placed on ice (a, b) or on the bench at room temperature (c, d) for different time periods (n=10) up to 8 hours.

A similar phenomenon was observed in serum samples; both inosine and hypoxanthine levels increase with increasing clotting time. While both hypoxanthine and inosine increased with clotting time at room temperature, after ~4 hrs. the inosine peaks started to diminish in intensity (Fig. S16). An analysis of intact plasma samples (without removing macromolecules) showed that inosine was not detectable at any of the blood processing times investigated (up to 144 hrs.) and hypoxanthine was detectable only at a much later processing time (>24 hrs.) (Figs. S17, S18). The suppression of inosine and hypoxanthine in intact samples potentially arises from their binding to copious proteins (60–80 g/L) present in the samples. It is well-known that metabolites including many amino acids and organic acids bind to plasma/serum proteins, which makes their peaks either invisible or significantly attenuated.^33–36 Hence, hypoxanthine and inosine will be less useful as biomarkers for assessing the preanalytical sample quality if intact plasma or serum is used for their measurement by NMR. This is significant considering that intact plasma/serum samples are widely used in the metabolomics field.

We also evaluated aqueous metabolites in the time series plasma samples using LC-MS/MS. To compare the changes in the levels of metabolites as a function of time, the relative concentrations of metabolites at the first time point (0 hr.) were set to 1 and for other time points, ratios of the relative concentrations of metabolites with respect to the concentrations at 0 hr. were calculated. Figure 5 shows plots of the changes in the levels for metabolites as a function of the time. It is striking to note that hypoxanthine is the most sensitive metabolite with respect to time among the quantitated aqueous metabolites (n=240), although, it is not surprising based on the results of NMR analysis described above. Further, in blood that stayed on ice, inosine is the next most prominent metabolite to increase significantly (Fig. 5a), though at a later time period. In blood that stayed at room temperature, oxidized glutathione appeared to increase and then decrease. However, glutathione is a notoriously unstable metabolite and blood samples were not pretreated with N-ethylmaleimide to prevent glutathione from oxidizing.²⁵ Hence, glutathione is unlikely to become a reliable sample quality marker. Thus overall, these results are in accordance with the results described above based on NMR analysis.

Figure 5:

Relative change in 240 LC-MS/MS derived aqueous human plasma metabolite levels when whole blood was placed on ice (a) or on the bench at room temperature (RT) (b) for different times prior to separating plasma. The relative change indicates the ratio between metabolite level when blood stayed on ice or on the bench at RT for time > 0 hr. and 0 hr.

Regarding other potential biomarkers of sample quality, lactate is often regarded as a biomarker of plasma sample quality. When blood processing is delayed, plasma lactate levels increase due to glycolysis in RBC, which is associated with glucose consumption at a rate of 5 to 7 % per hour.⁹ However, lactate and glucose are already present in both RBC and plasma and their concentrations vary widely between individuals. Further, lactate levels increase with exercise as well as under pathological conditions.³⁷ Hence, any increase in the lactate levels due to the delay in blood processing is often indistinguishable from physiological variations. Hence, neither lactate nor glucose can be a robust biomarker for plasma/serum quality. As an example, Figure S19 shows lactate and glucose levels at different time intervals for the same samples as shown in Figure 6; changes in the levels of glucose and lactate are in accordance with our previous study²⁶ and it is clear from the results that lactate is a late marker of poor sample quality and thus largely unsuitable for predicting the quality of plasma samples. We also investigated 2,3-BPG, another major blood metabolite produced in RBC through the Luebering–Rapoport pathway of glycolysis (Scheme S1). 2,3-BPG is critical for the supply of oxygen to all cells in the human body.³⁸ The results show that it does not contribute to altered plasma/serum metabolites during the time period for which the blood was on ice or on the bench. On the contrary, when whole blood was stored on ice, the levels of 2,3-BPG slightly increased with time, which is indicative of the active Luebering–Rapoport pathway metabolism in RBC (Fig. S20).

Figure 6:

Box and whisker plots of the concentrations of hypoxanthine and inosine in plasma as a function of the time and temperature prior to processing. The times and the respective median values and ranges for hypoxanthine and inosine levels (μM). See Table S3 for median and range values for each plot.

To further illustrate the utility of inosine and hypoxanthine as potential biomarkers of plasma/serum quality, we evaluated >2000 ¹H NMR spectra of plasma/serum that had been collected as part of other studies. The samples represent eight cohorts procured from the geographically distinct sites in the United States (Table S1). Figure S21 shows portions of NMR spectra for six of these cohorts along with highlighting of the characteristic peaks from inosine and hypoxanthine. The median levels of inosine and hypoxanthine were in the range 0 – 17.4 μM and 1.7 – 72.4 μM, respectively (Fig. 7). In comparison, the results from this study show that the levels of inosine were 0 μM and the levels of hypoxanthine were in the range 0 – 1.4 μM when blood stayed on ice or on the bench at room temperature for up to 4 hrs. (Fig. 6; Table S3). These results indicate that a majority of samples in the cohorts have alarmingly high levels of hypoxanthine and inosine, which signify that the blood was on ice or at room temperature for an unacceptably long time (perhaps up to several days) (Figs. 7 and S21). And importantly, the high levels of inosine indicate that blood samples very likely were kept at cold temperatures (Fig. 6). The cohort of serum samples also showed high levels of hypoxanthine and inosine, which indicates that the blood samples were very likely kept for a long clotting period prior to processing (Fig. S22). In one cohort of plasma samples high levels of ATP, ADP, and AMP were detected, which indicates hemolysis of blood (Fig. S23). Overall, these results indicate the gravity of the compliance challenges for sample procurement in metabolomics-centric studies. The challenge is largely unrecognized and underappreciated, which is aggravated by the lack of reliable biomarkers to validate or predict the sample quality. Hence, the use of both hypoxanthine and inosine as biomarkers for plasma sample quality as indicated here provides more detailed information on sample processing and thus sample quality.

Figure 7:

Box and whisker plots of concentrations of hypoxanthine and inosine in six cohorts of human plasma samples from different regions of the United States (see also Fig. S21). The median values and ranges of hypoxanthine and inosine concentrations (μM), respectively, are: (a) cohort 1: 72.4 (1.1, 249.4), 17.4 (0.0, 83.7); (b) cohort 2: 5.8 (0.0, 46.0), 0.6 (0.0, 15.1); (c) cohort 3: 6.1 (0.0, 34.8), 0.4 (0, 9.3); (d) cohort 4: 5.4 (0.0, 44.3), 2.4 (0.0, 15.8); ((e) cohort 5: 1.7 (0.0, 9.5), 0.0 (0.0, 3.2); and (f) cohort 6: 3.2 (0.0, 14.3), 0.9 (0.0, 6.9).

Hypoxanthine and inosine show two peaks in the region 8.2 to 8.4 ppm, which is generally devoid of interference from other metabolite peaks (Figs. 2, 3, S21). Further, each peak is a singlet as they arise from isolated hydrogen atoms from their purine ring moieties with no adjacent hydrogens to cause multiplicity due to spin-spin couplings (Scheme 1). Singlet peaks provide the best resolution and sensitivity compared to peak multiplets. Although one peak is sufficient for quantitation, the redundancy offers an opportunity for verification of proper identification as well as the flexibility to choose the peak that does not overlap with other metabolites for quantitation. For example, as shown in Figure S21, in contrast to the peak at 8.243 ppm the inosine peak at 8.348 ppm often overlaps with an unidentified peak and hence it is generally unsuitable for quantitation. Further, for accurate quantitation using NMR spectra, it is important to ensure that a sufficient relaxation delay is used or the peak area corrected for relaxation. There is more than one approach to make such correction. One approach is to obtain spectrum for the same or similar sample with recycle delay long enough to obtain fully relaxed peaks and use these peak areas to correct the area for the same peaks in spectra obtained using a shorter recycle delay. The other approach is to determine T₁ values and use them to make corrections to peak areas as described previously.²⁸ In the latter case, however, it is important to ensure that the same or similar type of sample is used for T₁ determination since T₁ values depend on sample composition and solvent. For example, the use of deuterated (D₂O) buffer causes longer relaxation times (up to a factor of two) compared to buffer prepared using water (H₂O) (Table S4) and the use of such values yields inaccurate concentrations.

The phenomenon of the accumulation of hypoxanthine in blood/plasma was first detected nearly a century ago, although the knowledge of its origin was unknown.^39–41 However, its utility as a robust biomarker for assessing plasma/serum sample quality in the metabolomics field, to date, has not been recognized widely. A few studies have reported its increased levels apart from many others in serum/plasma;^6,7,1,13,21 it was, however, suggested that metabolite profiles are stable when blood is placed on ice and processed within four hours.^6,7 On the contrary, we detected high levels of hypoxanthine and inosine in plasma even when blood stayed on ice (Fig. 6) and our results are in accordance with the anomalous dynamics of ATP in cold blood discovered recently,²⁶ which leads to the accumulation of hypoxanthine as well as inosine in plasma. As further illustration, Figure S24 shows the levels of ATP, ADP, AMP, IMP, inosine, and hypoxanthine in a typical untreated cold human blood as a function of time. It shows, as anticipated, that the total pool of metabolites associated with the dynamics of ATP remains the same at all time points with an average error of < 4.6%. Further, remarkably, these biomarkers are detectable by both NMR spectroscopy and MS, consistently, with high sensitivity and specificity. Their absolute concentrations being measurable by NMR using a single internal standard or even without the need for an internal standard^28,42 offers a simple avenue for the assessment of plasma/serum sample quality.

While the current study demonstrates the utility of hypoxanthine and inosine as biomarkers, the study also stresses the critical need to avoid the contamination of plasma/serum metabolite profiles due to the cellular metabolism, which may deleteriously affect the inferences of metabolomics studies. It should, however, be noted that hypoxanthine and/or inosine in plasma are reported as biomarkers for exercise,^43–46 liver transplantation, and numerous diseases including cardiac ischemia,⁴⁷ hypoxia,⁴⁸ xanthinuria,⁴⁹ cancer,⁵⁰ obstructive sleep apnea syndrome,⁵¹ and acute respiratory distress syndrome.⁵² In view of the accumulation of hypoxanthine and inosine even in healthy subjects as demonstrated in the current study (Fig. 7), it is critical that the confounding effects due to the delayed blood processing be accounted for evaluating their potential role under pathological conditions, if any. To achieve this, blood needs to be processed for plasma as soon as it is drawn and serum clotting times should be minimized. Alternatively, the cellular metabolism needs to be quenched using an organic solvent as we have demonstrated earlier.²⁶ Given that immediate blood processing can be generally challenging logistically, quenching the metabolism is a desirable option. As described in detail in our previous study,²⁶ quenching enables accurate measurement of labile metabolites such as redox coenzymes (NADH, NAD⁺, NADPH, NADP⁺), energy coenzymes (ATP, ADP, AMP), and antioxidants (GSH, GSSG) and a large number of other metabolites including IMP, hypoxanthine, inosine, glucose and lactate, whose levels are massively altered due to the metabolism of blood cells. Overall, the findings provide compelling evidence for a paradigm shift in sample harvesting protocols for blood metabolomics.

In conclusion, we describe the discovery of hypoxanthine and inosine as potential biomarkers to assess or validate the preanalytical quality of plasma/serum, a widely used biospecimen in the metabolomics field. These biomarkers are highly specific since they are generally not present in plasma/serum and they accumulate in blood as it ages either at room temperature or even at cold temperatures due to the anomalous dynamics of ATP in cold blood. They are highly sensitive since they are derived from ATP, which is one of the highly concentrated metabolites in human blood. Interestingly, our results also show that, based on plasma inosine levels, whether or not blood was kept at room temperature or on ice could be predicted. To the best of our knowledge these are the only biomarkers associated with the metabolism of blood cells that are excreted into plasma at high concentrations enough to detect early even by NMR spectroscopy. Such characteristics enable their detection routinely by both NMR and MS. The fact that these biomarkers are derived from a highly unstable and highly concentrated blood metabolite, ATP, gives support for their high sensitivity and specificity. Their surprisingly high levels found in sample cohorts procured based on metabolomics-centric protocols indicate the gravity of the compliance challenges for sample procurement. Henceforth, in addition to their concentrations, the characteristic region (8.2 to 8.4 ppm) of ¹H NMR spectra of plasma/serum that show hypoxanthine and inosine peaks should be used as an important visual yardstick to evaluate the quality of plasma/serum samples used in metabolomics studies on a routine basis. Both NMR and MS are widely used in blood metabolomics. While MS is highly sensitive, NMR is highly quantitative and reproducible. Hence, NMR is more suitable for discovering biomarkers and developing methods for plasma/serum preanalytical quality assessment. Since MS invariably uses protein precipitation prior to analyses, biomarkers developed using NMR, under similar conditions, can be directly translated for MS analyses. For this reason, in this study, we used NMR to develop and describe the biomarkers’ discovery and then demonstrated their utility even in MS analysis. Overall, the discovery of robust biomarkers provides new opportunities to address the grossly underrecognized and underappreciated challenge of sample quality in the metabolomics field.